Identifying Dissatisfied customers

50 %
50 %
Information about Identifying Dissatisfied customers

Published on November 21, 2016

Author: VamshiVennamaneni

Source: slideshare.net

1. Santander : Customer Satisfaction Chaitanya Kartheek , Sai Rajendra , Vamshi V Classifier Model  The 65 features selected by ‘SelectFrom Model’ are passed to AdaBoostClassifier for classification.  As the target variable is skewed , we had to balance the class in 80:20 ratio. We observed the results obtained after class balancing are much better.  The final classifier output is a majority vote of AdaBoost, Bagging and XGBoost models Problem: This is a classification problem for identifying dissatisfied customers from their banking experience Feature Selection  After all the data processing activities, we were left with 275+ variables. As it is difficult to interpret and also not computationally productive to run the model on all the 275 variables , we have performed feature selection using RandomForestClassifier.  After fitting a RandomForestClassifier on the training dataset, we were able to identify the important features but needed a concrete logic to select the features.  We have included all the features that are more than half of mean of all feature scores using SelectFromModel package Results & Evaluation: We have used AUC, Confusion matrix, Accuracy scores and Matthew’s coefficient as metrics to evaluate our model Th . Statistical Learning & Analytics Spring, 2016 Data and Preprocessing  We have been provided an anonymized dataset with 371variables.This is a challenge as we cannot use any prior knowledge to select or increase weights to any features  The target variable in the dataset is highly skewed ~ 96%  In this preprocessing phase, we have eliminated 30+ variables having zero standard deviation and removed columns that are exactly same as they do not contribute to the model Conclusion  The business opportunity for this project is to identify dissatisfied customers so that their concerns can be addressed. So, Specificity is key metric in this scenario and the, conventional metrics like accuracy scores, AUC can be misleading.  At a banker level, for a given customer, the top 5 features are critical and hence banker needs to check the health of these features and take suitable measures.

Add a comment

Related pages

Identifying latently dissatisfied customers and measures ...

Customer satisfaction continues to be an important topic in the financial services industry. However, there is an increasing awareness that customer satisfaction as ...
Read more

What Is the Difference Between a Dissatisfied Customer ...

Dissatisfied and angry customers arrive at their feelings from different circumstances, but neither is good for your business. Soothing unhappy customers might ...
Read more

6 Signs Of An Unhappy Customer (Checklist)

Identifying unhappy customers is a tough gig. At Client Heartbeat, we do it every day for 100’s of companies. So, what makes an unhappy customer?
Read more

10 Need-to-Know Customer Dissatisfaction Stats

10 Need-to-Know Customer ... of dissatisfied customers ... Today’s businesses need to take a critical look at how they identify and handle all their dissatisfied ...
Read more

Identifying dissatisfied customers and measures for ...

Official Full-Text Publication: Identifying dissatisfied customers and measures for dissatisfaction management on ResearchGate, the professional network for scientists.
Read more

Tell-Tale Signs Your Customer is Confused, Upset, or ...

Tell-Tale Signs Your Customer is Confused, Upset, or Dissatisfied Customer Service Training Resource July 19, 2013 — 7,577 views
Read more

3 Ways to Prevent, Identify and Engage Unhappy Customers ...

Here are three ways you can prevent, identify and engage unhappy customers. ... identify and engage unhappy customers: 1. ... Identify Situations That Require Your ...
Read more

Dealing With Unhappy Customers - Communication Skills from ...

Don't be afraid of dealing with unhappy customers - turn a bad experience into a good one. One of Tim's most important clients has just walked into his office ...
Read more

Prevent, Identify and Engage Unhappy Customers on Social ...

With the explosion of digital channels, online reputation management (as well as high quality customer service) is more critical than ever. Customer ...
Read more

Training and Education for Customer Service Professionals

Live Customer Service Training Events. Live Webinars for Customer Service Professionals. How to Control Your Emotions When Dealing With Customers
Read more